Optical Extraction of Oil Spills from Satellite Images under Different Sunglint Reflections.
Xiaobo Zhu,Yingcheng Lu,Jianqiang Liu,Weimin Ju,Jing Ding,Manchun Li,Ziyi Suo,Junnan Jiao,Lifeng Wang
DOI: https://doi.org/10.1109/tgrs.2022.3213590
IF: 8.2
2022-01-01
IEEE Transactions on Geoscience and Remote Sensing
Abstract:Optical remote sensing is applied in the identification, classification, and quantification of weathered oil spills. The automatic detection of oil spills through optical imaging is yet a challenge because various oils under different sunglint reflections have complex optical image characteristics. Generally, there are two types of weathered oil spills, namely, non-emulsified oil slicks (NEOS) and oil emulsions (OE), which show different image characteristics under various sunglint reflections. The Coastal Zone Imager (CZI) onboard China’s HaiYang-1C/D (HY-1C/D) satellites can provide multispectral images with high spatial resolution and wide coverage for operational monitoring of oil spills. In this study, we applied an adaptive dynamic detector incorporating a built-in oil–water mixture distribution classifier, specifically for different sunglint reflections, to automatically extract oil spills from CZI images. The spatial heterogeneity distribution of various oil spills could be quantified using a novel separability index, and then, the optimal oil–water segmentation proportion and scale could be obtained. Oil spills are discriminated and extracted under different sunglint reflections, with the variable scale detector implemented by tiling sliding windows of classifiers on detection images, from which respective volumes are derived with lower uncertainties. This approach also uses spatial and spectral ancillary information to improve weathered oil extraction confidence. The results show stable variable-scale extraction accuracies of approximately 90% and 80% for NEOS and EO, respectively. Therefore, the spatio–spectral–distribution comprehensive feature provides a new approach for the automatic extraction of oil spills from optical remote sensing images.